ML Image Compression Latent Space Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression methods struggle to balance compression ratio and visual accuracy, particularly in lossy compression techniques where the trade-off between file size and image quality is not user-specific.
Innovation Solution
A machine learning (ML) compression model is used to identify compressible portions of image data, select appropriate ML compression models, and generate ML-compressed representations, which include text and vector combinations. These representations are then used to create a compressed image data file that can be decompressed using an ML decompression model, allowing for user-specific compression based on visual perception.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If lossy compression methods are used to reduce file size, then compression ratio is improved, but visual accuracy of the reconstructed image deteriorates
Solution Approach 1:
The patent applies parameter changes by transforming image data from pixel space to latent space using machine learning models. The compression process encodes image data into a compressed latent representation with reduced parameters, while the decompression model reconstructs the image by decoding these parameters back to pixel space, achieving both compression and acceptable visual quality
Solution Approach 2:
The patent replaces traditional mechanical/image-processing compression algorithms with machine learning-based compression models. Instead of using conventional compression techniques that operate on pixel values directly, the system uses trained neural networks to learn efficient representations of image data, enabling better compression ratios while maintaining visual accuracy
2Manufacturing precision
If user-specific compression is implemented to improve visual accuracy, then image quality is improved, but device complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-training compression and decompression models during an offline phase. These pre-trained models capture user-specific visual perception characteristics, allowing the system to apply optimized compression without requiring complex real-time processing during actual compression operations
Solution Approach 2:
The patent introduces an intermediary latent space representation that bridges the gap between compressed data and reconstructed images. This latent space serves as a mediator where compression occurs, allowing the system to manage complexity by operating in this intermediate representation rather than directly manipulating pixel data
Data Source
AI summary
A method includes obtaining image data, identifying a machine learning-compressible (ML-compressible) portion of the image data, and determining a location of the ML-compressible portion within the image data. The method also includes selecting, from a plurality of ML compression models, an ML compression model for the ML-compressible portion based on an image content thereof, and generating, based on the ML-compressible portion and by the ML compression model, an ML-compressed representation of the ML-compressible portion. The method further includes generating a compressed image data file that includes the ML-compressed representation and the location of the ML-compressible portion, and outputting the compressed image data file. The compressed image data file is configured to cause an ML decompression model corresponding to the ML compression model to generate a reconstruction of the ML-compressible portion of the image data based on the ML-compressed representation.